13 binary signals
Heading hierarchy, FAQ block, lists, structured data, a key takeaway, an intro under 80 words, a named author and six more, each present or absent.
Be Retrieved, the third rung
Audit a page against published thresholds
Most content scores are a letter grade and a vague note to write better. This one names the band your page fell outside, and by how much.
Key takeaway
Heading density below 0.5 per 100 words leaves nothing to chunk on. Above 3.0, headings stop meaning anything. Both ends fail, which is the whole argument for bands over a score: "write better" becomes a number that is out of range.
What gets checked
Every page is fetched server-side and parsed, so the audit sees what a crawler sees rather than what a browser renders.
11 structural bands
Paragraphs 30 to 90 words, list ratio 0.10 to 0.40, heading density 0.5 to 3.0 per 100 words, and eight more, all published as numbers.
Signals for what your site is
A lead-generation site is also checked for comparison content, testimonials and case studies; an ecommerce site for product schema and pricing.
Technical state alongside
Status code, indexability, robots directive, canonical, title, meta, H1 count, schema types and link counts, all from the same pass.
Why bands rather than a score
A composite grade tells you a page is weak. A band tells you which number to change.
Macro, document architecture
Heading depth, section count, heading density and whether a table of contents exists. This decides whether an engine finds the passage.
Meso, information chunking
Average and 90th-percentile paragraph length, list ratio and chunk variance. A page averaging 60 words can still hide a 400-word wall.
Micro, visual emphasis
Emphasis density, blockquotes, code density and image-to-text ratio. Too little gives an extractor no anchor, too much flattens the signal.
Bands, not thresholds to maximise
Each band has a floor and a ceiling because both ends fail. Too few headings is no structure, too many and every heading means less.
A readable diff on every change
Each crawl stores a snapshot, so the audit reports which fields moved and a score change is always attributable to an edit.
Rolls up to the rung
The average audit score across your pages is the Be Retrieved rung. One page failing is a task, the average moving is a programme.
What it changes
The point of a numeric band is that it ends the argument about whether a page is good enough.
It separates writing from structure
Most pages that fail extraction are well written and badly structured. A 400-word paragraph with no headings is not a quality problem.
It makes a rewrite checkable before publishing
The bands are computable on unpublished content, so you never publish, wait a fortnight and infer whether structure was the problem.
It feeds a measured experiment
A page change becomes a change event with a field-level diff, and the prompts it should have moved become the treatment group.
Against a content score
The difference is whether the output names a number you can change.
Published numeric bands with floors and ceilings
TrustData
Google Analytics 4
Other tools
Server-side fetch, sees what a crawler sees
TrustData
Google Analytics 4
Other tools
SometimesSignals differ by site type
TrustData
Google Analytics 4
Other tools
Field-level diff between crawls
TrustData
Google Analytics 4
Other tools
Rolls up into a readiness rung
TrustData
Google Analytics 4
Other tools
| Feature | TrustData | Google Analytics 4 | Other tools |
|---|---|---|---|
| Published numeric bands with floors and ceilings | |||
| Server-side fetch, sees what a crawler sees | Sometimes | ||
| Signals differ by site type | |||
| Field-level diff between crawls | |||
| Rolls up into a readiness rung |
FAQ
Direct answers.
It scores whether a page can be parsed, chunked and quoted by an AI engine, which is a different question from whether it ranks. TrustData fetches each page server-side, parses the HTML, and checks 13 binary signals along with 11 structural bands covering document architecture, information chunking and visual emphasis. The output is the specific bands the page fell outside, not a grade.
Because both ends fail. Heading density below 0.5 per 100 words means the page has no structure to chunk on; above 3.0 means headings stop carrying meaning. Paragraphs under 30 words fragment an idea across chunks; over 90 they bury the answer. A list ratio under 0.10 means nothing is enumerable, over 0.40 means the prose has become a bullet dump with no connective reasoning. Every band has a floor and a ceiling for that reason.
They derive from published experimental work on generative-engine extraction, and they are deliberately treated as a first pass rather than a law. The intention is to regress them against observed citation rates on real properties, so the bands tighten as evidence accumulates. Until then they are documented rather than hidden, which is the difference between a threshold you can argue with and a black-box score you cannot.
Yes. Beyond the shared signals, a lead-generation site is checked for comparison content, testimonials, case studies, use cases and social proof, because those are what an engine looks for when answering an evaluation question about a service. An e-commerce site is checked for product schema, visible pricing and reviews instead. Scoring both against the same list would penalise one for lacking something it should not have.
The average audit score across your pages is the Be Retrieved rung, which is third. It sits above Get Found, because a page a crawler cannot reach cannot be extracted whatever its structure, and below Be Trusted, because a perfectly structured page still needs corroboration to be cited with confidence. If Get Found is failing, the audit is not your bottleneck yet, and the ladder will say so by marking the lower rung as the focus.
14-day free trial
Audit a page against the bands
14-day free trial. Page audits are part of AI Visibility, included in every plan.